The SNES sound chip plays samples in its own compressed format, BRR (Bit Rate Reduction — a 9-byte block per 16 samples). wav2brr encodes an ordinary PCM .wav into a .brr you load at runtime with audioLoadSample(). It is the zero-config path for one-shot sound effects.

What goes in, what comes out

In: a PCM .wav — 8- or 16-bit, mono or stereo (stereo is downmixed).

Out: a raw .brr (its size is always a multiple of 9 bytes). Default name is input.brr alongside the input.

The flags you will actually use

Flag	Meaning
--loop START END	loop between sample indices START..END (default: one-shot, no loop)
-v	verbose — prints the size breakdown, a sample-rate warning, and a ready-to-paste load line

How you actually use it

One-shot SFX — automatic. Drop a .wav in the example's res/ and .incbin the matching .brr in an ASM file; the build's %.brr: %.wav rule runs wav2brr for you. No command to type:

res/blip.wav ──(make)──► res/blip.brr ──.incbin──► audioLoadSample()

Looping samples — by hand. A sustained instrument or a looping ambience needs loop points, which only you know, so build it once and commit the .brr:

wav2brr --loop 512 2048 instrument.wav instrument.brr

Run with -v to see the block count and the exact audioLoadSample() call to paste.

Gotchas worth knowing up front

Keep the source at or below 32 kHz. That is the DSP's output ceiling; a higher-rate WAV is accepted but plays sharp (a note higher than authored). -v warns when it happens. The pitch you hear is set at playback by the DSP pitch value, relative to the sample's rate — see SNES Sound Programming Guide.
Looping bypasses the auto-rule. The zero-config path makes one-shots only; anything with --loop is a hand-built, committed asset.
Deterministic and shared. wav2brr uses the same BRR encoder as smconv — tracker modules to SNESMOD soundbanks, so a sample sounds identical whether it is a standalone SFX or baked into a soundbank (golden-tested via make test-tools).

See it in practice

SFX from WAV — one-shots auto-generated from res/*.wav.
soundboard — the audio v2 engine, driven entirely from C — committed .brr blobs, .incbin'd and played with per-voice volume/pan/pitch.

For music (as opposed to individual samples), you want smconv — tracker modules to SNESMOD soundbanks and its tracker-module workflow instead.